Existing integrated assessment models for climate change use conventional economic approaches to implement the social representations. This paper argues that these economic approaches would be inappropriate if economists were rational in the sense they assume their agents to be. A straightforward stochastic, cost-benefit analysis shows that conditions of application of any model should be identified and investigated unless (a) it is extremely expensive to determine whether they are satisfied or (b) the policy analyst is convinced that any policies implied by his model are better than anything else that could be done even if the model is inapplicable in the prevailing circumstances. Well known theorems and results from the economics literature are then shown to imply that the conditions of application of the economic models used in integrated assessment analysis are so implausible as to render those models useless in that (or probably any) policy context. A constructive alternative programme for integrated assessment is then proposed.